Proefschrift

154 During the technical investigation phase of our research, we operationalised the Glass Box framework by creating an implementation concept as an example to prove that the framework is actionable. After value elicitation, deriving norms and requirements is the next step in the interpretation stage of the Glass Box framework. These requirements will feed into the observation stage as observable elements (criteria) to monitor and verify. We created an implementation of a pre- and post-flight procedure of an autonomous drone as an example using Coloured Petri Nets (CPNs) as modelling language. The implementation concept shows that it is possible to set criteria in the pre-flight process and to evaluate these criteria post-flight. During flight, the drone itself is treated as a black box of which the internal logic is not accessible. This way the users do not need technical skills to understand the internal workings of an autonomous drone, but still can monitor and oversee the use of the autonomous system based on observable norms. A review stage is required after deployment as an accountability process of which findings should feed back into the interpretation stage for a next deployment of an autonomous system and thereby close the loop between the stages. Finally, we returned to the conceptual investigation phase of our research and describe a toy example in which we apply the Five-Point Systems feedback loop to the case of Autonomous Weapon Systems. The purpose of the feedback system is to ensure that the lessons and recommendations from the review stage will be incorporated in the interpretation stage before deployment of an Autonomous Weapon System in a next iteration. Results of this research are the delineation of accountability, responsibility and human oversight concepts which adds to the current body of literature. Also, the Comprehensive Human Oversight Framework and implementation concept of the Glass Box framework show that criteria for Human Oversight can be identified, represented and validated. This leads to a proper allocation of accountability in the decision-making of the deployment of an Autonomous Weapon System and it might be possible to attribute responsibility for the actions taken by the weapon system by identifying the supervisor of these actions. This thereby contributes to decreasing the likelihood of unintended consequences in the deployment of Autonomous Weapon Systems. SUMMARY

RkJQdWJsaXNoZXIy MjY0ODMw